Fivethrityeight had a few neat articles on how correlated each state was to the national vote and how that's accounted for in the model. So it basically uses similar states and adjusts for demographics to get a +/- from the national average.
I think the flipping of Ohio and Iowa is saying that even though Ohio and Iowa are 4-5 points more republican then the national average Clinton's 4-5 point lead is enough to counter act this.
|